You are on page 1of 15

Applied Mathematical Modelling 89 (2021) 333–347

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

Time-dependent interaction between superstructure, raft and


layered cross-anisotropic viscoelastic saturated soils
Zhi Yong Ai∗, Zhong Hao Chu, Yi Chong Cheng
Department of Geotechnical Engineering, Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, College
of Civil Engineering, Tongji University, Shanghai 200092, PR China

a r t i c l e i n f o a b s t r a c t

Article history: By introducing the elastic-viscoelastic correspondence principle and the integral transform
Received 14 July 2019 technique, the extended precise integral solution to Biot’s consolidation equation is
Revised 24 June 2020
derived for cross-anisotropic viscoelastic saturated soils based on the Merchant model.
Accepted 5 July 2020
A displacement-time solution for a slowly changing load is obtained based on the above
Available online 4 August 2020
solution in the Laplace transformed domain, which is taken as the kernel function
Keywords: for the boundary element method (BEM). The Mindlin plate model is established by
Cross-anisotropic 8-node isoparametric finite element method (FEM), and the substructure condensation
Layered saturated soils technique is employed to couple the stiffness matrices of the superstructure and the
Viscoelasticity plate. With the BEM-FEM coupling method, a semi-analytical and semi-numerical solution
Soil-structure interaction is proposed for the interaction between layered cross-anisotropic viscoelastic saturated
Superstructure stiffness soils and raft foundation considering the stiffness contributions of the superstructure.
Time-dependent behavior Numerical examples are performed to study the influences of viscoelastic parame-
ters, cross-anisotropic parameters, raft thickness and superstructure stiffness on the
time-dependent behavior of the settlement and bending moment for the raft.
© 2020 Elsevier Inc. All rights reserved.

1. Introduction

In practical projects, the basic function of the raft foundation is to transfer the load from the superstructure to soils or
rocks. Therefore, the stress distribution and deformation of the superstructure, foundation and soils are closely related, and
it is necessary to regard them as a whole for analysis. Nonetheless, due to the limitation of computational theories and
means, most of the current plate-soil interaction studies [1–12] only consider the load transferred from the superstructure,
and neglect its stiffness contributions to the interactive plate-soil system.
Meyerhof [13] first proposed the concept of superstructure-plate-soil interaction. After that, the following quasi-static
analysis methods for this problem have been developed. The first type is the equivalent replacement method [14], which
assumes that the superstructure is replaced by a beam system whose stiffness is several times that of the foundation beam,
and is connected to the foundation by hinged columns at two ends; however, such a scheme is only suitable for qualitative
analysis. The second type is the holistic modeling solution [15,16], which directly models and solves the superstructure
together with the foundation; however, for complex structures, the calculation can only be implemented by means of
pure numerical methods such as the FEM, which requires high computational and storage performance of the computer.


Corresponding author.
E-mail address: zhiyongai@tongji.edu.cn (Z.Y. Ai).

https://doi.org/10.1016/j.apm.2020.07.018
0307-904X/© 2020 Elsevier Inc. All rights reserved.
334 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

The third type is the substructure solution [17,18], which divides the superstructure into a plurality of subsystems, and the
stiffness and load of the superstructure can be condensed onto the raft foundation to carry out the calculation. Compared
with the holistic FEM, the substructure method has less demand for computer memory and performs high efficiency in
calculation. Nevertheless, the above studies [14-18] cannot simulate the time-dependent behavior of state parameters, such
as the settlements and bending moments of the raft.
Very little research on the time-dependent behavior of quasi-static interaction between the raft-superstructure system
and saturated soils have been reported. Vlladkar et al. [19] presented a three-dimensional (3D) viscoelastic FEM-formulation
to study the frame-foundation-soil interaction, taking into account the stress/strain-time response of soils. Nasri & Magnan
[20] studied the time effect of consolidation in elastic soils and ideal elastoplastic soils on the behavior of 3D frame-raft-
soil system. Wang et al. [21] researched the interaction of the superstructure, raft and isotropic viscoelastic saturated soils
based on the substructure method and thin plate theory. Overall, the correlative studies are still not perfect enough since
they ignored the anisotropic characteristics of soils. Hence, it is necessary to establish a more reasonable theory on the
interaction analysis of the superstructure, raft and soils in the time domain by comprehensively considering the anisotropy,
stratification and rheological characteristics of saturated soft soils.
In this study, the state vector equation for axisymmetric Biot’s consolidation in cross-anisotropic saturated soils is de-
duced by taking the Laplace integral transform [22] on the time variable t and the Hankel integral transform [22] on the
coordinate variable r. The extended precise integration method [23] is used to solve the state vector equation. Based on
the elastic-viscoelastic correspondence principle [24] and the Merchant model, the obtained solution is extended to vis-
coelastic soils. To introduce the coupled BEM-FEM technique, we divide the plate-soil interface with boundary elements
in the same way as the division of the finite elements on the plate. And the distributed loads on the elements of plate-
soil interface are transformed into point loads through interpolation functions for convenience of analysis, which are the
equivalent nodal forces. Then a displacement-time solution for a slowly changing point load [9] is obtained in the Laplace
transformed domain and taken as the kernel function for the BEM. The Mindlin plate is modeled by the FEM, and the sub-
structure technique is further introduced to take the stiffness contributions of the superstructure into account. By extending
the coupled BEM-FEM method in related studies [6,25], a semi-analytical and semi-numerical solution is established for the
time-dependent interaction analysis between the raft-superstructure system and layered saturated soils. Numerical exam-
ples are given to verify the accuracy of this solution and to investigate the time-dependent effect of superstructure-plate-soil
interaction.

2. The precise integration solution to the consolidation of cross-anisotropic viscoelastic saturated soils

2.1. The precise integration solution of cross-anisotropic elastic saturated soils

In this paper, we assume that cross-anisotropic planes are parallel to free surface, and thus z-axis is the symmetry
axis of soil system [26]. Then the equilibrium differential equations without body forces for axisymmetric problems in the
cylindrical coordinate system are:

∂ σr ∂ σrz σr − σθ
+ + = 0, (1a)
∂r ∂z r
∂ σz ∂ σrz σrz
+ + = 0, (1b)
∂z ∂r r
where σ r , σ θ , σ z stand for the normal stress components in the r, θ , z directions, respectively; σ rz is the shear stress
component in the r-z plane.
According to the principle of effective stress, we have:

σs = σ  s − σw , (2)

where σ s = [σ r ,σ θ ,σ z ,σ rz and σ     T
]T s = [σ r ,σ θ ,σ z ,σ rz ] denote the total stress vector and the effective stress vector, re-
spectively; σ w = [σ w ,σ w ,σ w ,0]T is the vector of excess pore fluid pressure (pressure as positive).
The constitutive relation of a cross-anisotropic medium can be written as:
⎡ ⎤
c11 c12 c13 0
 ⎢c12 c11 c13 0⎥
σs=⎣ ε,
0⎦ s
(3)
c13 c13 c33
0 0 0 c44

where εs = [ ∂∂urr , urr , ∂∂uzz , ∂∂uzr + ∂∂urz ]T , ur and uz represent the displacement components in the r,z directions, re-
spectively; c11 = λζ (1 − ζ νv2h ), c12 = λζ (νh + ζ νv2h ), c13 = λζ ν vh (1 + ν h ), c33 = λ(1 − νh2 ), c44 = Gv , in which λ =
Ev /[(1 + νh )(1 − νh − 2ζ νv2h )],ζ = Eh /Ev ; Eh , Ev are the horizontal Young’s modulus and vertical Young’s modulus, respec-
tively, and Gv is the shear modulus on the vertical plane; ν h and νvh stand for the Poisson’s ratios characterizing the lateral
strain due to the stress acting parallel and normally to the plane, respectively.
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 335

Fig. 1. Merchant viscoelastic model.

The seepage continuous equation is [27]:


 2

∂e 1 ∂ σw 1 ∂ σw ∂ 2 σw
= kh + + k v , (4)
∂ t γw ∂ r2 r ∂r ∂ z2
where e = ∂∂urr + urr + ∂∂uzr stands for the volumetric strain; kh and kv represent the coefficients of permeability in the horizontal
direction and vertical direction, respectively; γ w is the unit weight of pore water.
On the basis of Darcy’s law, the total flow within the time from 0 to t in the z direction can be expressed as [23]:
t
k v ∂ σw
Qz = dt. (5)
0 γw ∂ z
By taking the Laplace and Hankel integral transforms on the time variable t and coordinate variable r, respectively, Ai
and Cheng [23] derived a state vector equation for the consolidation of layered elastic soils for non-axisymmetric problems
in the cylindrical coordinate system. Herein, with the similar method and based on the above five equations, we obtain:
d
W1 (z ) = H1 · W1 (z ), (6)
dz
   
0 b2 ξ b3 ξ b4 0 0 b5 ξ 2 0 0
A1 D1
where H1 = , and A1 = −ξ 0 0 , B1 = 0 b1 b1 , D1 = 0 0 0 , in which b1 = 1
c33 ,
B1 −AT1
0 0 0 0 b1 ϑ 0 0 Fz
2
c11 c33 −c13
c13 −c33 sγ k
, Fz = k w , ϑ = c1 + sγhw ξ 2 ; W1 (z) = [V1 ,U1 ]T , in which V1 = [σ̄rz1 , σ̄z0 , σ̄w0 ] and
c13 1
b2 = c33 , b3 = c33 , b4 = c44 , b5 = c33 v 33
U1 = [ū1r , ū0z , Q̄z0 ] stand for the generalized stress vector and displacement vector in the transformed domain, respectively;
s and ξ are the transform parameters with respect to the variables t and r, respectively; σ̄rz1 , σ̄z0 , σ̄w0 , ū1r , ū0z , Q̄z0 are the
corresponding forms of the original variables in the transformed domain and the superscript indicates the order of Hankel
transform.
Assume that the surface of the soils is free and completely drained, and the foundation base is fixed and impervious;
that is, the generalized stress vector of the surface and the generalized displacement vector of the bottom are both zero
vectors. For Eq. (6), if the boundary conditions of the two ends and the external load conditions of the soils are given, the
extended precise integration method proposed by Ai & Cheng [23] can be utilized to calculate the displacement and stress
of any point in the soils.

2.2. The Merchant model

The Merchant model [28] is a classical rheological model composed of an elastomer and a Kelvin viscoelastic body con-
nected in series, the structure and parameters of which are shown in Fig. 1, and its constitutive equation is given as:
η1 d σ E0 E1 η E dε
σ+ = ε+ 1 0 , (7)
E0 + E1 dt E0 + E1 E0 + E1 dt
where η1 is the viscosity coefficient, E0 and E1 stand for the elastic moduli of Hooke body H0 and H1 , respectively; σ and ε
are the total stress and strain, respectively.
The Merchant model can synthetically reflect the rheological properties of soft soils, so we employ it to simulate satu-
rated soils in this study.

2.3. The elastic-viscoelastic correspondence principle

Lee [24] found that the form of the basic governing equation of viscoelasticity body is the same as that of the linear
elasticity theory in the Laplace transformed domain, which means the solutions of viscoelastic problems can be acquired by
replacing elastic constants of the linear elastic solutions with viscoelastic parameters in the transformed domain. And the
actual solutions in the physical domain can be obtained through numerical inversions. By means of the elastic-viscoelastic
correspondence principle, the consolidation solution of elastic saturated soils can be converted to the corresponding solution
for viscoelastic saturated soils.
336 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

As for an isotropic body, the stress-strain relationship is only determined by the elastic modulus under the premise that
the Poisson’s ratio is a constant. Hence the correspondence between viscoelasticity and elasticity can be achieved in the
Laplace transformed domain through replacing the elastic modulus in the elastic solutions with the reciprocal of flexibility
coefficient V˜ (s ) = ε˜ (s )/σ˜ (s ), here the superscript ‘~’ stands for the variable in the transformed domain. With the assumption
that the initial strain and initial stress are zero, then after taking the Laplace transform to Eq. (7), the flexibility coefficient
for the Merchant model in the transformed domain is acquired as follow:
1 1
V˜ (s )= + . (8)
E0 E 1 + s η1
However, the stress-strain relation depends on several parameters in the cross-anisotropic viscoelastic media, so the
corresponding flexibility coefficients of Eh , Ev and Gv in the Laplace transformed domain need to be introduced, which can
be defined as:
ε˜h (s ) ˜ ε˜v (s ) ˜ ε˜g (s )
V˜h (s )= , V (s )= , V (s )= . (9)
σ˜ h (s ) v σ˜ v (s ) g σ˜ g (s )
To expand the application of elastic-viscoelastic correspondence principle in this case, we assume that the viscosity co-
efficient and elastic moduli between the vertical and lateral directions have a constant proportional relation. Thus, the flex-
ibility coefficients can be expressed as:
1 1
ς · V˜g (s ) = ζ · V˜h (s ) = V˜v (s ) = + , (10)
E0v E 1 v + s η1 v

where ς =V˜v (s )/V˜g (s ) and ζ = V˜v (s )/V˜h (s ) = E0h /E0v = E1h /E1v = η1h /η1v are the proportional coefficients of parameters be-
tween shear and vertical directions as well as lateral and vertical directions, respectively; E0v , E0h , E1v , E1h , η1v and η1h
denote the parameters in the lateral and vertical directions of the Merchant model, respectively. Although this is just a
special case, the related research [29] proves that it is quite applicable to practical engineering.

3. The superstructure-raft-soil interaction

To introduce the coupled BEM-FEM technique, we divide the plate-soil interface with boundary elements in the same
way as the division of the finite elements on the plate. The distributed loads on the elements of plate-soil interface are
transformed into point loads through interpolation functions for convenience of analysis, which are the equivalent nodal
forces.

3.1. The relationship between variable loads and soil displacement

For the raft on saturated soils, the pressure on the contact element slowly changes in pace with the plate-soil interaction;
thus, the solution for a constant load in Section 2 cannot be directly applied in this case. Therefore, a time-displacement
solution under a slowing changing load is further derived in this section based on the previously mentioned solution.
When a force imposed at point α (xα ,yα ,0) on the surface of soils varies slowly with time, it can be expressed in an
integral form as follows:
t
∂ p( α , τ )
p( α , t ) = p( α , 0 ) + dτ , (11)
0 ∂τ
where p(α , t) denotes the force imposed at point α at time t.
The vertical displacement uz (α ,β , t) and excess pore pressure σ w (α ,β , t) at point β (xβ ,yβ ,zβ ) at time t induced by the
load p(α , τ ) acting at point α at time τ (which remains unchanged afterwards) can be expressed as:

uz (α , β , t ) = ϕ1 (α , β , t − τ ) p(α , τ ), (12a)

σw (α , β , t ) = ϕ2 (α , β , t − τ ) p(α , τ ), (12b)

where ϕ 1 (α ,β , t − τ ) and ϕ 2 (α ,β , t − τ ) are the flexibility coefficients standing for the vertical displacement and excess
pore pressure at point β (xβ ,yβ ,zβ ) at time t caused by a constant unit point load acting at point α at time τ , respectively,
which can be obtained from the solution in Section 2.
Substituting Eq. (11) into Eqs. (12a) and (12b) we have:
t
∂ p( α , τ )
uz ( α , β , t ) = ϕ1 ( α , β , t ) p( α , 0 ) + ϕ1 ( α , β , t − τ ) dτ . (13a)
0 ∂τ
t
∂ p( α , τ )
σw ( α , β , t ) = ϕ 2 ( α , β , t ) p ( α , 0 ) + ϕ2 ( α , β , t − τ ) dτ . (13b)
0 ∂τ
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 337

Fig. 2. The eight-node Mindlin plate element.

To solve Eqs. (13a) and (13b), we need to introduce the convolution formula [22]:
t
f1 (t ) ∗ f2 (t ) = f1 (τ ) f2 (t − τ )dτ (14)
0

where f1 (t) and f2 (t) are functions related to time t; ∗ is the convolution symbol.
The convolution has the following property:

˜[ f1 (t ) ∗ f2 (t )] = ˜[ f1 (t )] · ˜[ f2 (t )] (15)


where ˜[ f (t )] represents the Laplace transformation of f(t) relative to time t.
Take the Laplace transform of Eqs. (13a) and (13b), and combine Eqs. (14) and (15), we can obtain:

u˜z (α , β , s ) = R˜1 (α , β , s ) p˜ (α , s ), (16a)

σ˜ w (α , β , s) = R˜2 (α , β , s) p˜ (α , s ), (16b)

where R˜1 (α , β , s ) = sϕ˜ 1 (α , β , s ), and R˜2 (α , β , s ) = sϕ˜ 2 (α , β , s ); the superscript ‘~’ represents the variable in the transformed
domain.
Through Eqs. (16a) and (16b), the vertical displacement and excess pore pressure induced by the slowly-varying point
load in the Laplace transformed domain can be obtained.
When the vertical displacement u˜z (S, β , s ) and the excess pore pressure σ˜ w (S, β , s ) of point β are induced by the loads
distributed over the area S, the integral expressions of Eqs. (16a) and (16b) are:

u˜z (S, β , s ) = R˜1 (α , β , s ) p˜ (α , s )dS, (17a)
S

σ˜ w (S, β , s) = R˜2 (α , β , s ) p˜ (α , s )dS. (17b)
S

3.2. The substructure method for the analysis of superstructure-raft system

In this study, the three-degree-of-freedom and eight-node Mindlin plate isoparametric element is used to build the finite
element model, as shown in Fig. 2. Mx and My are the bending moments in the x and y directions, respectively, and Mxy
is the torque in the xy plane; θ x and θ y are the rotation angles in the x, y directions, respectively. Considering that the
stiffness matrix of the plate does not change with time, the global equilibrium equation of a Mindlin plate assembled by
element stiffness equations can be written as [30]:

KrUr (t ) = Fr (t ) − Qs (t ), (18)
where Kr is the global stiffness matrix of the plate, Ur (t), Fr (t) and Qs (t) are the vectors of the displacement, the external
force and the subgrade reaction force at time t, respectively.
The substructure method is employed to calculate the interaction between the superstructure and raft. As shown in Fig. 3,
taking a structural layer in the space frame structure as the substructure, according to the FEM, the equilibrium equation is
given as:

K f U = P, (19)
where Kf , U and P represent the stiffness matrix, the displacement vector and the load vector of a substructure, respectively.
338 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

Fig. 3. The space frame and its substructure.

Fig. 4. The condensation of substructures.

The above matrix equation is partitioned according to the inner node and boundary node, i.e.:
  
Kii Kib Ui Pi
· = , (20)
Kbi Kbb Ub Pb

where Ui and Ub stand for the displacement vectors of the inner node and boundary node, respectively; Pi and Pb stand for
the load vectors of the inner node and boundary node, respectively.
After the elimination of Ui in Eq. (20), we obtain:
KbUb = Sb , (21a)
where,

Kb = Kbb − Kbi Kii−1 Kib , (21b)

Sb = Pb − Kbi Kii−1 Pi , (21c)


in which Kb and Sb denote the equivalent boundary stiffness matrix and equivalent boundary load vector, respectively.
The degrees of freedom of the inner nodes are eliminated by Eq. (21), and thereby the condensation from the substruc-
ture to the boundary node is realized.
For a structural system consisting of m substructures, the condensation process is shown in Fig. 4. Thus, when the equiv-
alent boundary stiffness matrices and the equivalent boundary load vectors of n − 1 substructures are incorporated into the
equilibrium equation of the nth substructure, we have:
  
Kii( ) + Kb( ) Kib( ) Ui( ) Pi( ) + Sb( )
n n−1 n n n n−1

(n ) (n ) · (n ) = . (22)
Pb( ) − Rb( )
n n
Kbi Kbb Ub

where Rb(n ) stands for the boundary reaction force vector of the nth substructure.
According to the recursive relationship of Eq. (22), the substructures can be condensed from the top to the bottom until
the mth substructure, that is:
Kb( )Ub( ) = Sb( ) − Rb( ) ,
m m m m
(23)

where Kb(m ) and Sb(m ) can also be obtained through the above condensation theory.
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 339

Similarly, Eq. (23) can be coupled with the stiffness matrix equation of the plate. The global stiffness matrix and global
equilibrium equation of the Mindlin plate considering the stiffness of the superstructure are derived below.
For convenience of explanation, taking into account the displacement compatibility and force equilibrium of the nodes
on a Mindlin plate jointed with the column on the mth substructure, Eq. (18) can be rewritten as:
   
Ub( ) F + Rb( )
m m
Kcc Kcd Qc
· = c − , (24)
Kdc Kdd Ud Fd Qd

where the subscript ‘c’and ‘d’ denote the nodes jointed with the column and other nodes of the plate, respectively.
Combining Eq. (23) with Eq. (24), the new equilibrium equation can be acquired as follows:
   
Kcc + Kb( ) Ub( ) F + Sb( )
m m m
Kcd Qc
· = c − . (25)
Kdc Kdd Ud Fd Qd

Eq. (25) can be simplified as an expression similar to formula (18):


K Ur (t ) = F  (t ) − Qs (t ). (26)
Eq. (26) is the global equilibrium equation of the plate-superstructure system. Compared with Eq. (18), and are K F  ( t)
the global stiffness matrix of the Mindlin plate and the external force vector of nodes at time t after the condensation of
superstructures, respectively.

3.3. The raft-soil coupling equation

The plate-soil interface is supposed to be frictionless and divided by boundary elements, which are entirely identical
with the finite elements of the plate. Based on the assumption that the vertical point load p(α , t) on an arbitrary point α in
an element has an interpolated relationship with the vertical tractions on its eight nodes at time t [31], we get:

p(α , t ) = NP r( ) (t ),
e
(27)
(e )
where N represents the vector of interpolation function, Pr (t ) denotes the vector of vertical tractions on the eight nodes
of element e.
Taking the Laplace transform of Eq. (27) and then inserting the transformed equation into Eqs. (17a) and (17b), the
vertical displacement u˜z (S(e ) , β , s ) and excess pore pressure σ˜ w (S(e ) , β , s ) of any point β caused by the vertical tractions on
element e in the transformed domain can be written as:
   
u˜z S(e ) , β , s = R˜1 (α , β , s )N dS(e ) P˜r( ) (s ),
e
(28a)
S (e )
   
σ˜ w S(e) , β , s = R˜2 (α , β , s )N dS(e ) P˜r( ) (s ),
e
(28b)
S (e )

where S(e) is the area covered by the element e.


Moreover, the vertical displacement u˜z (Sr , β , s ) and excess pore pressure σ˜ w (Sr , β , s ) caused by the vertical tractions on
the total boundary elements of the interface are:
m 
 
u˜z (Sr , β , s ) = R˜1 (α , β , s )N dS(e ) P˜r( ) (s ),
e
(29a)
e=1 S (e )

m 
 
σ˜ w (Sr , β , s) = R˜2 (α , β , s )N dS(e ) P˜r( ) (s ),
e
(29b)
S (e )
e=1

where Sr is the area covered by the plate, and m stands for the number of the elements.
Through Eq. (29a), when β represents all of the soil-raft boundary nodes, the total vertical displacements on the nodes
of the boundary elements can all be calculated, and thus the relationship between the traction vector and the vertical
displacement vector is further obtained as:
˜ r (s ) = H˜ P˜r (s ),
W (30)
where W ˜ r (s ) and P˜r (s ) stand for the total vector of the vertical displacement and nodal traction vector composed of all the
nodes on the soil-raft interface in the Laplace transformed domain, respectively; and H˜ is the flexibility matrix obtained
from Eq. (29a).
Similarly, when β represents the nodes used for the calculation of excess pore pressure, the relationship between the
traction vector and the excess pore pressure vector can also be obtained as:
σ˜ w (s ) = Y˜ P˜r (s ), (31)
where Y˜ is the flexibility matrix obtained from Eq. (29b).
340 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

Table 1
Main calculated parameters for the layered viscoelastic saturated soils.

Layer E0v /MPa E1v /MPa η1v /MPa · s ζ ς ν h ( = ν vh ) kv (= kh )/cm · s−1 h/m

1 21.0 8.4 1E7 1.0 0.357 0.40 1E-6 3.0


2 14.4 4.8 1E7 1.0 0.345 0.45 1E-6 6.0
3 10.0 2.5 1E7 1.0 0.345 0.45 1E-7 10.0
4 63.0 25.2 1E7 1.0 0.357 0.40 1E-6 +∞

According to the FEM, the equivalent nodal force vector to the vertical forces of element e is:

Qr( ) (t ) = N T p(α , t )dS(e ) ,
e
(32)
S (e )

where Qr(e ) (t ) is the vector of the vertical equivalent nodal force on element e.
Combining Eq. (27) with Eq. (32), we have:

Qr( ) (t ) = T (e ) Pr( ) (t ),
e e
(33)

where T (e ) = S(e) N T N dS(e ) represents the transformation matrix of element e.
The total transformation matrix equation in the Laplace transformed domain can be established by Eq. (33):
Q˜r (s ) = T P˜r (s ), (34)
where Q˜r (s ) and T denote the total vertical equivalent nodal force vector in the transformed domain and the total transfor-
mation matrix assembled by T(e) , respectively.
Combining Eq. (30) and Eq. (34), we acquire:
Q˜r (s ) = G˜W
˜ r (s ), (35)
where G˜ = T H˜ −1 .
Note that Eq. (35) only demonstrates the relationship between the equivalent nodal force and displacement in the vertical
direction; however, each node in the Mindlin plate element has three degrees of freedoms. To achieve the compatibility of
the finite elements of the raft with the boundary elements of the soil surface, we assume that the deformation of soils is
not related to the rotation angle of the raft, and thus the expansion form of Eq. (35) after introducing zero values of the
bending moments and rotation angles is:
Q˜s (s ) = G˜ sW
˜ rs (s ), (36)
where Q˜s (s ) and W˜ rs (s ) are the expanded vectors of Q˜r (s )and W
˜ r (s ) in the transformed domain, respectively; G˜ s is the
stiffness matrix obtained from G˜ . The columns and rows corresponding with the variables of bending moments and rotation
angles in Eq. (36) are zero vectors.
The continuity condition of the displacement on the raft-soil interface is:
Ur (t ) = Wrs (t ). (37)
where Wrs (t) is the expanded total vertical displacement vector of the BEM in the time domain.
Taking the Laplace transforms of Eqs. (26) and (37), and then combining them with Eq. (36), we can obtain:
 
K  + G˜ s U˜r (s ) = F˜ (s ), (38)

in which U˜r (s ) and F˜ (s ) are the corresponding vectors in the transformed domain.
Eq. (38) is the coupling equation of the superstructure-raft-soil interaction in the Laplace transformed domain. The real
displacement solution in the time domain can be acquired by solving the generalized displacement vector in Eq. (38) and
taking the Laplace inverse transform. Then the boundary forces of the plate can be obtained by Eq. (30). Since the boundary
forces are known, the excess pore pressure in soil layers can be obtained by Eq. (31). Futhermore, the bending moments in
the raft can be calculated through the FEM.

4. Numerical examples

4.1. Verification

To verify the accuracy of the present method in this study, we make a comparison of the results acquired in this study
with that proposed by Wang et al. [21] under the same situation. A rectangular raft of length L= 26m, width B = 14m,
thickness hr = 0.4m, elastic modulus Er =30 0 0 0MPa and Poisson’s ratio ν r = 0.2 is supported by a four-layered viscoelastic
saturated soil system, the main calculated parameters of which are listed in Table 1. The superstructure is an 8-story space
frame with a story-height of 3 m. There are 6 and 3 spans along the length and width direction of the raft, respectively, and
each span is 4 m. The section dimension of the frame beam and column are 0.6m × 0.3m and 0.5m × 0.5m, respectively; the
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 341

Fig. 5. Comparison of the vertical displacement of the central point on the raft bottom with the existing reference [21].

elastic modulus and Poisson’s ratio of them are 32,500 MPa and 0.2, respectively. The load on each floor is q0 = 20kN/m2 . It
can be clearly found from Fig. 5 that the vertical displacement of the center point O on the bottom of the raft calculated in
this study is in good agreement with that in the existing reference [21], which demonstrates the reliability of the presented
approach.

4.2. The influence of soil viscoelasticity and cross-anisotropy

To investigate the effects of soil viscoelasticity and cross-anisotropy on the interaction, six cases are designed for cal-
culation and analysis, in which the calculated depth of soils is h = 40m; kh = kv = 1 × 10−6 cm/s and ν h = ν vh = 0.35.
For cases 1, 2, 3, 4 and 5: ς = 0.37, E0 h /E1 h =E0 v /E1 v =2, E1v =15MPa. For cases 1, 2 and 3: η1v = 1 × 107 MPa·s,
5 × 107 MPa·s and 1 × 108 MPa·s, respectively; ζ = E0 h /E0v = 0.5; the equivalent final moduli are Ev∗ = 10MPa, Eh∗ = 5MPa,
here 1/Ev∗ = 1/E0 v + 1/E1 v , 1/Eh∗ = 1/E0h + 1/E1h , which can reflect the compressive deformation properties of soils to a cer-
tain extent. For cases 4 and 5: ζ =1.0 and 2.0, respectively; η1v = 1 × 108 MPa·s. For case 6, the cross-anisotropic elastic soil
model is applied for comparison with the above five viscoelastic cases, in which Ev = Ev∗ = 10MPa and Eh = Eh∗ = 5MPa. The
parameters of the plate and the superstructure in the six cases are identical with those in Section 4.1. The compressibility of
pore water is not taken into consideration for the numerical examples in this study. The dimensionless factors of time, dis-
E∗ k E∗ u
placement, bending moment and excess pore pressure are defined as τ = BLvγwv t, w∗ = Bvq z , M∗ = q MB2 , and σw∗ = nσs wq (ns = 8
0 0 0
is the story number) in this section, respectively.
Figs. 6 and 7 depict the time-dependent behavior of deflection and bending moment respectively at the center point
O on the raft bottom in cases 1, 2, 3 and 6. It can be observed that the final values obtained in different cases are equal
under the same equivalent final modulus, but the time-dependent behavior of plate-soil interaction is more significant in
the viscoelastic model. At any time during consolidation, the internal force and deformation of the plate on viscoelastic soils
are less than that on elastic soils. With the increase of η1v , the curves gradually become less steep, which indicates the
deformation and bending moment of the plate are smaller at any time. When the viscous coefficient increases, the bonding
strength of the combined water increases and the viscosity increases monotonically, which makes it more difficult for the
soil particles to overcome the sliding resistance. Therefore, it takes a longer time for the internal force and deformation in
the raft to reach a stable value.
Fig. 8 illustrates the time-dependent behavior of the excess pore pressure at a depth of B under the corner point of raft
bottom. It can be found from Fig. 8 that as the viscosity coefficient gets larger, the Mandel-Cryer effect [32,33] occurs earlier,
in which the excess pore pressure increases initially and then decreases with time. However, the variation is not significant.
The similar trend can also be seen in Ref. [34].
Figs. 9 and 10 reflect the influence of ζ ( = E0h /E0v ) on the time-variation of deflection and bending moment at point
O, respectively. As can be seen, as ζ becomes larger, the deflection at point O is smaller at any time during consolidation.
This is because in the case where E0v and E1v are constant, the increase of E0h and E1h reduce the horizontal deformation
of the raft, which leads to the decrease of the total settlement of the foundation. In addition, the deflection-time curves of
the three cases are approximately parallel. Results shown in Fig. 10 demonstrate that the bending moment of point O also
decreases with the increase of ζ .
The influence of ζ on the excess pore pressure at a depth of B under the corner point of raft bottom is shown in Fig. 11.
From Fig. 11, the Mandel-Cryer effect is found in the case of ζ = 0.5 and ζ = 1, and is more significant in the former
342 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

Fig. 6. Influence of η1v on the vertical displacement at the central point on the raft bottom versus time factor.

Fig. 7. Influence of η1v on the bending moment at the central point on the raft bottom versus time factor.

Fig. 8. Influence of η1v on the excess pore pressure at a depth of B under the corner point of raft bottom versus time factor.
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 343

Fig. 9. Influence of ζ (= E0 h /E0v ) on the vertical displacement at the central point on the raft bottom versus time factor.

Fig. 10. Influence of ζ (= E0 h /E0v ) on the bending moment at the central point on the raft bottom versus time factor.

Fig. 11. Influence of ζ (= E0 h /E0v ) on the excess pore pressure at a depth of B under the corner point of raft bottom versus time factor.
344 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

Table 2
Main parameters for the three-layered viscoelastic saturated soils.

Layer E0v /MPa E1v /MPa η1v /MPa · s ν vh νh ζ ς kv (= kh )/cm · s−1 h/m

1 25 6 5E7 0.30 0.30 2.0 0.40 1E-6 10.0


2 18 8 1E8 0.35 0.30 1.5 0.37 2E-6 20.0
3 10 3 1E8 0.40 0.35 1.0 0.30 5E-6 20.0

Fig. 12. The frame layout of each floor.

Fig. 13. Influence of raft thickness on the differential settlement versus time factor.

case. While there is no Mandel-Cryer effect in the case of ζ = 2 because the increasing of Eh accelerates the transfer of
the effective stress to the internal of soils. Besides, the soil cross-anisotropy coefficient has a noticeable effect on the excess
pore pressure during consolidation process. The similar observation can also be found in Ref. [35].

4.3. The influence of raft thickness

This example is used to study the effect of the raft thickness (hr ), and the parameters of the 3-layered cross-anisotropic
viscoelastic saturated soils are listed in Table 2. The superstructure contains 5 storeys, and the frame layout of each floor is
shown in Fig. 12. The section dimension of the frame beams and columns are 0.5m × 0.3m and 0.5m × 0.5m, respectively.
The elastic modulus of the beams, columns and raft is 32,500 MPa, and the Poisson’s ratio of them is 0.2. The width of the
square raft is B1 = 10m, and the load transferred from the superstructure to the foundation plate is q. The raft thickness is
0.4 m, 0.5 m, 0.6 m and 0.7 m in the four cases of this section, respectively.
In Sections 4.3 and 4.4, the dimensionless factors of time, displacement, differential settlement, bending moment and
(1−ν̄ )Ē k̄v (uz (O )−uz (C ))
, M∗ = M2 and σw∗ = σqw
Ē uz Ē
excess pore fluid pressure are defined as τ = vh 1v t, w∗ = B1v q , w∗ = 1v B q
(1+ν̄vh )(1−2ν̄vh )B21 γw 1 1 qB1
   
respectively, where ν̄vh = 3i=1 ν (vh )i hi /hs , Ē1v = 3i=1 E(1v )i hi /hs , k̄v = 3i=1 kvi hi /hs and hs = 3i=1 hi . The subscript i stands
for the coefficient of Layer i. O and C represent the central point and the corner point on the raft bottom, respectively.
As depicted in Fig. 13, when hr is increased from 0.4 m to 0.7 m, the differential settlement between the central point
and the corner point is continuously reduced. It is because that as the raft thickness increases, the equivalent stiffness of the
foundation increases correspondingly, and thus the performance of the raft to resist uneven settlement enhances. However,
as hr continues to increase, the decreasing extent of the differential settlement is also reduced. Fig. 14 illustrates that the
bending moment of the central point is positively correlated with hr . The reason is that when the relative stiffness between
the foundation and soils is larger, the base reaction force tends to be concentrated toward two ends, and thus the bending
moment of the central point is larger.
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 345

Fig. 14. Influence of raft thickness on the bending moment at the central point on the raft bottom versus time factor.

Fig. 15. Influence of raft thickness on the excess pore pressure at a depth of 2B1 under the corner point of raft bottom versus time factor.

Fig. 15 reveals the influence of raft thickness on the excess pore pressure at a depth of 2B1 under the corner point of
raft bottom. With the increase of raft thickness, the excess pore pressure also increases, but the change is slight. Besides,
the Mandel-Cryer effect fails to occur because the ζ value of Layer 1 is relative larger, as discussed in Section 4.2.

4.4. The influence of superstructure stiffness

The influence of superstructure stiffness is investigated in this example, where hr is selected as 0.5 m. The model param-
eters are the same as those in Section 4.3 except for the number of storeys of the framed structure. Four cases with 0, 5,
10 and 15 storeys are taken into account. Figs. 16, 17 and 18 depict the time-dependent behavior of differential settlement,
central bending moment and excess pore pressure, respectively. It is concluded that with the increase of the storey number,
both the differential settlement and the bending moment reduce, and the time-dependent behavior tends to be less signif-
icant. However, when the storey number increases, the decrease extent of the differential settlement and bending moment
in the plate is less than before. From the perspective of mechanism analysis, the superstructure with a certain rigidity has
a restraining effect on the foundation, so that the overall bending of the raft can be avoided. It should be pointed out that
the superstructure stiffness and the exerted load do not change with time in this example, so the calculated results in this
section only reflect the bending moment and deformation varying with time after the completion of construction. Besides,
it can be concluded from Fig. 18 that the influence of superstructure stiffness on the excess pore pressure is not significant.
346 Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347

Fig. 16. Influence of superstructure stiffness on the differential settlement versus time factor.

Fig. 17. Influence of superstructure stiffness on the bending moment at the central point on the raft bottom versus time factor.

Fig. 18. Influence of superstructure stiffness on the excess pore pressure at a depth of 2B1 under the corner point of raft bottom versus time factor.
Z.Y. Ai, Z.H. Chu and Y.C. Cheng / Applied Mathematical Modelling 89 (2021) 333–347 347

5. Conclusions

The extended precise integration solution to Biot’s consolidation for cross-anisotropic saturated elastic soils is expanded
to cross-anisotropic saturated viscoelastic soils based on the Merchant model, which is taken as the kernel function for the
BEM. The Mindlin plate is modeled by the FEM and the substructure technique is further applied. With BEM-FEM coupling
theory, an analysis method of superstructure-raft-soil interaction in the time domain is proposed. The reliability of the
presented method is proved by comparing the results in this study with the existing results. Numerical examples show that
in contrast with the elastic models, the time-dependent behavior of interaction is more significant with the increase of the
viscosity coefficient in the viscoelastic models. When the vertical elastic modulus of soils is constant, both the settlement
and the bending moment at the central point on the raft bottom decrease as the horizontal elastic modulus increases. The
uneven settlement in the raft is alleviated with the increase of the raft thickness and superstructure stiffness. The effect of
soil cross-anisotropy parameter on the excess pore pressure is noticeable, while the influences of soil viscosity coefficient,
plate thickness and superstructure stiffness are slight.

Acknowledgments

This research is supported by the National Natural Science Foundation of China (Grant No. 41672275).

References

[1] R.A. Fraser, L.J. Wardle, Numerical-analysis of rectangular rafts on layered foundations, Geotechnique 26 (4) (1976) 613–630.
[2] J.C. Small, B.Q. Zhang, Finite layer analysis of the behavior of a raft on a consolidating soil, Int. J. Numer. Anal. Methods Geomech. 18 (4) (1994)
237–251.
[3] E.S. Melerski, Numerical modelling of elastic interaction between circular rafts and cross-anisotropic media, Comput. Struct. 64 (1-4) (1997) 567–578.
[4] L. Sadecka, A finite/infinite element analysis of thick plate on a layered foundation, Comput. Struct. 76 (5) (20 0 0) 603–610.
[5] Y.H. Wang, Y.K. Cheung, Plate on cross-anisotropic foundation analyzed by the finite element method, Comput. Geotech. 28 (1) (2001) 37–54.
[6] Y.H. Wang, L.G. Tham, Y. Tsui, Z.Q Yue, Plate on layered foundation analyzed by a semi-analytical and semi-numerical method, Comput. Geotech. 30
(5) (2003) 409–418.
[7] A.A. Katebi, A. Khojasteh, M. Rahimian, R.Y.S Pak, Axisymmetric interaction of a rigid disc with a transversely isotropic half-space, Int. J. Numer. Anal.
Methods Geomech. 34 (12) (2010) 1211–1236.
[8] S.L. Chen, Y. Abousleiman, Time-dependent behaviour of a rigid foundation on a transversely isotropic soil layer, Int. J. Numer. Anal. Methods Geomech.
34 (9) (2010) 937–952.
[9] Z.Y. Ai, Y.C. Cheng, G.J Cao, A quasistatic analysis of a plate on consolidating layered soils by analytical layer-element/finite element method coupling,
Int. J. Numer. Anal. Methods Geomech. 38 (13) (2014) 1362–1380.
[10] Z.Y. Ai, Y.D. Hu, A coupled BEM-ALEM approach for analysis of elastic thin plates on multilayered soil with anisotropic permeability, Eng. Anal. Bound.
Elem. 53 (2015) 40–45.
[11] Z.Y. Ai, Y.F. Zhang, The analysis of a rigid rectangular plate on a transversely isotropic multilayered medium, Appl. Math. Model. 39 (20) (2015)
6085–6102.
[12] Z.Y. Ai, Z. Cao, J.J. Mu, B.K Shi, Elastic thin plate resting on saturated multilayered soils with anisotropic permeability and elastic superstrata, Int. J.
Geomech. 18 (10) (2018) 04018137.
[13] G.G. Meyerhof, The settlement analysis of building frames, Struct. Eng. 25 (9) (1947) 369–409.
[14] G.G Meyerhof, Some recent foundation research and its application to design, Struct. Eng. 31 (6) (1953) 151–167.
[15] B.Q. Zhang, J.C. Small, in: Finite Layer Analysis of Soil-Raft-Structure Interaction, XIII CIMSTF, New Delhi, India, 1994, pp. 587–590.
[16] V.S. Almeida, J.B de Paiva, A mixed BEM-FEM formulation for layered soil-superstructure interaction, Eng. Anal. Bound. Elem. 28 (9) (2004) 1111–1121.
[17] J.M. Haddadin, Mats and combined footings-analysis by the finite element method, Proc. ACI 68 (12) (1971) 945–949.
[18] S.J. Hain, I.K. Lee, Rational analysis of raft foundation, J.e Geotechn. Eng. Div. ASCE 100 (GT7) (1974) 843–860.
[19] M.N. Vlladkar, G. Ranjan, R.P Sharma, Soil-structure interaction in the time domain, Comput. Struct. 46 (3) (1993) 429–442.
[20] V. Nasri, J.P. Magnan, Effect of soil consolidation on space frame-raft-soil interaction, J. Struct. Eng. 123 (11) (1997) 1528–1534.
[21] J. Wang, J. Chen, J Pei, Interaction between superstructure and layered visco-elastic foundation considering consolidation and rheology of soil, J. Build.
Struct. 23 (4) (2002) 59–64 in Chinese.
[22] I.N. Sneddon, The Use of Integral Transform, McGraw-Hill, New York, 1972.
[23] Z.Y. Ai, Y.C. Cheng, Extended precise integration method for consolidation of transversely isotropic poroelastic layered media, Comput. Math. Appl. 68
(12) (2014) 1806–1818.
[24] E.H. Lee, Stress analysis in visco-elastic bodies, Q. Appl. Math. 13 (1955) 183–190.
[25] J.J. Mandal, D.P. Ghosh, Prediction of elastic settlement of rectangular raft foundation—A coupled FE-BE approach, Int. J. Numer. Anal. Methods Ge-
omech. 23 (3) (1999) 263–273.
[26] A. Khojasteh, M. Rahimian, M Eskandari, Three-dimensional dynamic Green’s functions in transversely isotropic tri-materials, Appl. Math. Model. 37
(5) (2013) 3164–3180.
[27] S.J. Singh, R. Kumar, S Rani, Consolidation of a poroelastic half-space with anisotropic permeability and compressible constituents by axisymmetric
surface loading, J. Earth Syst. Sci. 118 (5) (2009) 563–574.
[28] R.E. Gibson, K.Y. Lo, A theory of consolidation for soils exhibiting secondary compression, Acta Polytech. Scand. (1961) 296: Chapter 10.
[29] X.Z. Liu, H.H. Zhu, Back analysis on staged construction in transversely isotropic viscoelastic soil and its application to geotechnical engineering, Chin.
J. Geotech. Eng. 24 (1) (2002) 89–92 in Chinese.
[30] Y.K. Cheung, O.C. Zienkiewicz, Plates and tanks on elastic foundation—An application of finite element method, Int. J. Solids Struct. 1 (4) (1965)
451–461.
[31] O.C. Zienkiewicz, R.L. Taylor, The Finite Element Method (4th), McGraw-Hill, London, 1991.
[32] J Mandel, Consolidation Des Sols, Geotechnique 3 (7) (1953) 287–299.
[33] C.W. Cryer, A comparison of the three-dimensional consolidation theories of Biot and Terzaghi, Q. J. Mech. Appl. Math. 16 (4) (1963) 401–412.
[34] Z.Y. Ai, Y.Z. Zhao, X.Y. Song, J.J Mu, Multi-dimensional consolidation analysis of transversely isotropic viscoelastic saturated soils, Eng. Geol. 253 (2019)
1–13.
[35] Y.C. Cheng, Z.Y. Ai, Consolidation analysis of transversely isotropic layered saturated soils in the Cartesian coordinate system by extended precise
integration method, Appl. Math. Model. 40 (4) (2016) 2692–2704.

You might also like